Block Diagram Reduction and Mason's Gain Formula for GATE: Transfer Functions with Solved Examples
Block diagram reduction and Mason's gain formula for GATE: transfer functions, feedback loops, reduction rules, signal flow graphs, worked examples and practice.
In the open loop vs closed loop post we saw why feedback is worth the trouble. Now it's time to handle it on paper. Real control systems are drawn as block diagrams with several loops, and before you can say anything about speed, error or stability, you need to boil the whole diagram down to one transfer function.
There are two ways to do that: block diagram reduction and Mason's gain formula on a signal flow graph. GATE asks one or the other almost every year in both EE and EC, and these are marks you can collect quickly once the rules are clear.
A quick recap: the transfer function
The transfer function of a system is the ratio of the Laplace transform of its output to that of its input, with all initial conditions zero:
For example, an RC low-pass circuit with the output taken across the capacitor has
The values of $s$ that make the denominator zero are the poles, and those that make the numerator zero are the zeros. Poles decide how the system responds, which is why we keep coming back to the denominator.
The one formula you must know: a feedback loop
Put a block $G$ in the forward path and a block $H$ in the feedback path, with the feedback subtracted. The error is $E = R - HC$ and the output is $C = GE$. Eliminate $E$ and you get
If the feedback is added instead of subtracted, the sign in the denominator flips to $1 - GH$. The expression $1 + G(s)H(s) = 0$ is called the characteristic equation, and its roots are the closed-loop poles. Keep that in mind for the stability posts to come.
The rules of block diagram reduction
Almost every reduction uses the same few moves:
Situation Replace it with ------ Blocks in series, $G1$ then $G2$ one block $G1 G2$ Blocks in parallel, outputs added one block $G1 + G2$ Negative feedback loop one block $\dfrac{G}{1 + GH}$ Moving a summing point from after $G$ to before it divide the other input by $G$ Moving a summing point from before $G$ to after it multiply the other input by $G$ Moving a take-off point from after $G$ to before it multiply the branch by $G$ Moving a take-off point from before $G$ to after it divide the branch by $G$
The moving rules all come from one idea: after the move, every signal must still be exactly what it was before.
Worked example: two loops
Here's a typical GATE diagram. The inner loop $H2$ wraps around $G2$, and the outer loop $H1$ goes all the way back to the input:
Step 1: close the inner loop. $G2$ with feedback $H2$ becomes
Step 2: combine with $G1$ in series.
Step 3: close the outer loop with feedback $H1$:
Always work from the innermost loop outwards. If loops overlap so that you can't close one without disturbing another, move a summing point or take-off point first to untangle them.
Signal flow graphs and Mason's gain formula
For diagrams with many interlocking loops, reduction gets messy. A signal flow graph draws the same system as nodes (signals) joined by arrows (gains), and Mason's formula gives the answer in one go:
where
$Pk$ is the gain of the $k$-th forward path from input to output, $\Delta = 1 - (\text{sum of all loop gains}) + (\text{sum of products of every pair of non-touching loops}) - (\text{sum of products of every three non-touching loops}) + \cdots$ $\Deltak$ is $\Delta$ worked out again using only the loops that don't touch the $k$-th forward path.
Here's the signal flow graph of Fig. 1:
Now apply the formula.
One forward path: $P1 = G1 G2$. Two loops: $L1 = -G2 H2$ and $L2 = -G1 G2 H1$. Both loops pass through $G2$, so they touch, and there are no non-touching pairs. That gives $\Delta = 1 + G2 H2 + G1 G2 H1$. Both loops touch the forward path, so $\Delta1 = 1$.
It's the same answer as the reduction, which is a good way to check yourself in the exam.
When loops don't touch
Mason's formula really earns its keep when loops are separate. Take a chain $G1 \to G2 \to G3$ with a local loop $-H1$ around $G1$ and another, $-H3$, around $G3$. The two loops share no node, so
and
Forgetting that product term, $G1 H1 G3 H3$, is the most common mistake in Mason's formula questions.
What the gain does to the response
Here's why all this matters. Take the simplest loop: unity feedback around $G(s) = K/(s + 1)$. Its closed-loop transfer function is
so the closed-loop pole sits at $s = -(1 + K)$, and the step response settles at $K/(1 + K)$. Raising $K$ makes the system faster and brings the output closer to the input:
That's feedback doing its job. The price, which we'll meet in higher-order systems, is that too much gain can push the system towards oscillation.
Common traps in GATE questions
These are the mistakes I see most often.
Sign of the feedback. Negative feedback gives $1 + GH$ in the denominator, and positive feedback gives $1 - GH$. Check the sign at the summing point. Moving points the wrong way. Moving a summing point ahead of a block divides by $G$. Moving it after multiplies. Say to yourself: the signals mustn't change. Missing non-touching products in Δ. Two loops that share no node contribute their product to $\Delta$. Wrong Δk. $\Deltak$ keeps only the loops that don't touch that forward path. If every loop touches it, $\Deltak = 1$. Counting a loop twice. A loop is a closed path that visits no node more than once. Starting it from a different node doesn't make it a new loop. Initial conditions. A transfer function assumes they're all zero. If a question gives initial conditions, you need the full Laplace solution. Non-unity feedback in steady state. With $H \ne 1$, the output settles at $G/(1 + GH)$ evaluated at $s = 0$, not at 1.
Practice questions
Give yourself about two minutes each.
Find the closed-loop transfer function for $G(s) = \dfrac{10}{s + 2}$ with unity negative feedback. Two blocks, $G1 = \dfrac{1}{s}$ and $G2 = 2$, are in parallel with their outputs added. Find the equivalent block. A summing point after a block $G$ is moved to before it. What happens to the other input to that summing point? A signal flow graph has one forward path of gain 6 and two loops of gain $-2$ and $-3$. The loops don't touch each other, and both touch the forward path. Find the transfer function. Write the characteristic equation for unity feedback around $G(s) = \dfrac{K}{s(s + 4)}$. For unity feedback around $G(s) = \dfrac{K}{s + 1}$, what value of $K$ makes the step response settle at 0.95?
Answers
$\dfrac{10}{s + 2 + 10} = \dfrac{10}{s + 12}$. $\dfrac{1}{s} + 2 = \dfrac{1 + 2s}{s}$. It must be divided by $G$, so that the output is unchanged. $\Delta = 1 - (-2 - 3) + (-2)(-3) = 1 + 5 + 6 = 12$, and $\Delta1 = 1$, so $T = 6/12 = 0.5$. $1 + \dfrac{K}{s(s + 4)} = 0$, which gives $s^2 + 4s + K = 0$. $\dfrac{K}{1 + K} = 0.95$ gives $K = 19$.
If question 4 slowed you down, write out $\Delta$ term by term, with signs, before you simplify. That habit saves marks every time. Next time we'll look at how these systems respond over time: first and second order responses, damping, overshoot and settling time. Tell me in the comments which question took you longest.